Artificial neural networks and deep learning.ppt

Artificial neural networks and deep learning.ppt

1. CSE 634
Data Mining Techniques
Presentation on Neural Network
Jalal Mahmud ( 105241140)
Hyung-Yeon, Gu(104985928)
Course Teacher : Prof. Anita Wasilewska
State University of New York at Stony Brook

2. References
Data Mining Concept and Techniques (Chapter 7.5)
[Jiawei Han, Micheline Kamber/Morgan Kaufman
Publishers2002]
Professor Anita Wasilewska’s lecture note
www.cs.vu.nl/~elena/slides03/nn_1light.ppt
Xin Yao Evolving Artificial Neural Networks
http://www.cs.bham.ac.uk/~xin/papers/published_iproc_sep99.pdf
informatics.indiana.edu/larryy/talks/S4.MattI.EANN.ppt
 www.cs.appstate.edu/~can/classes/
5100/Presentations/DataMining1.ppt
www.comp.nus.edu.sg/~cs6211/slides/blondie24.ppt
www.public.asu.edu/~svadrevu/UMD/ThesisTalk.ppt
www.ctrl.cinvestav.mx/~yuw/file/afnn1_nnintro.PPT

3. Overview
Basics of Neural Network
Advanced Features of Neural Network
Applications I-II
Summary

4. Basics of Neural Network
What is a Neural Network
Neural Network Classifier
Data Normalization
Neuron and bias of a neuron
Single Layer Feed Forward
Limitation
Multi Layer Feed Forward
Back propagation

5. Neural Networks
What is a Neural Network?
Similarity with biological network
Fundamental processing elements of a neural network
is a neuron
1.Receives inputs from other source
2.Combines them in someway
3.Performs a generally nonlinear operation on the
result
4.Outputs the final result
•Biologically motivated approach to
machine learning

6. Similarity with Biological Network
•Fundamental processing element of a
neural network is a neuron
•A human brain has 100 billion neurons
•An ant brain has 250,000 neurons

7. Synapses,
the basis of learning and
memory

8. Neural Network
Neural Network is a set of connected
INPUT/OUTPUT UNITS, where each
connection has a WEIGHT associated with it.
Neural Network learning is also called
CONNECTIONIST learning due to the connections
between units.
It is a case of SUPERVISED, INDUCTIVE or
CLASSIFICATION learning.

9. Neural Network
Neural Network learns by adjusting the
weights so as to be able to correctly classify
the training data and hence, after testing
phase, to classify unknown data.
Neural Network needs long time for training.
Neural Network has a high tolerance to noisy
and incomplete data

10. Neural Network Classifier
Input: Classification data
It contains classification attribute
Data is divided, as in any classification problem.
[Training data and Testing data]
All data must be normalized.
(i.e. all values of attributes in the database are changed to
contain values in the internal [0,1] or[-1,1])
Neural Network can work with data in the range of (0,1) or (-1,1)
Two basic normalization techniques
[1] Max-Min normalization
[2] Decimal Scaling normalization

11. Data Normalization
AnewAnewAnew
AA
Av
v min_)min_max_(
minmax
min
' 



[1] Max- Min normalization formula is as follows:
[minA, maxA , the minimun and maximum values of the attribute A
max-min normalization maps a value v of A to v’ in the range
{new_minA, new_maxA} ]

12. Example of Max-Min
Normalization
AnewAnewAnew
AA
Av
v min_)min_max_(
minmax
min
' 



Max- Min normalization formula
Example: We want to normalize data to range of the interval [0,1].
We put: new_max A= 1, new_minA =0.
Say, max A was 100 and min A was 20 ( That means maximum and
minimum values for the attribute ).
Now, if v = 40 ( If for this particular pattern , attribute value is 40 ), v’
will be calculated as , v’ = (40-20) x (1-0) / (100-20) + 0
=> v’ = 20 x 1/80
=> v’ = 0.4

13. Decimal Scaling Normalization
[2]Decimal Scaling Normalization
Normalization by decimal scaling normalizes by moving the decimal point of values of attribute A.
j
v
v
10
'
Here j is the smallest integer such that max|v’|<1.
Example :
A – values range from -986 to 917. Max |v| = 986.
v = -986 normalize to v’ = -986/1000 = -0.986

14. One Neuron as a
Network
Here x1 and x2 are normalized attribute value of data.
y is the output of the neuron , i.e the class label.

x1 and x2 values multiplied by weight values w1 and w2 are input to the neuron x.
Value of x1 is multiplied by a weight w1 and values of x2 is multiplied by a weight w2.
Given that
•w1 = 0.5 and w2 = 0.5
•Say value of x1 is 0.3 and value of x2 is 0.8,
•So, weighted sum is :
•sum= w1 x x1 + w2 x x2 = 0.5 x 0.3 + 0.5 x 0.8 = 0.55


15. One Neuron as a Network
•The neuron receives the weighted sum as input and calculates the
output as a function of input as follows :
•y = f(x) , where f(x) is defined as
•f(x) = 0 { when x< 0.5 }
•f(x) = 1 { when x >= 0.5 }
•For our example, x ( weighted sum ) is 0.55, so y = 1 ,
•That means corresponding input attribute values are classified in class 1.
•If for another input values , x = 0.45 , then f(x) = 0,
•so we could conclude that input values are classified to class 0.


16. Bias of a Neuron

We need the bias value to be added to the weighted
sum ∑wixi so that we can transform it from the origin.
v = ∑wixi + b, here b is the bias
x1-x2=0
x1-x2= 1
x1
x2
x1-x2= -1

17. Bias as extra input
Input
Attribute
values
weights
Summing function
Activation
function
v

Output
class
y
x
1
x
2
x
m
w
2
w
m
W1


)(
w
0
x
0
= +1
bw
xwv j
m
j
j



0
0

18. Neuron with Activation
The neuron is the basic information processing unit of a
NN. It consists of:
1 A set of links, describing the neuron inputs, with
weights W
1
, W
2
, …, W
m
2. An adder function (linear combiner) for computing the
weighted sum of the inputs (real numbers):
3 Activation function : for limiting the amplitude of the
neuron output.



m
1
jj xwu
j
) (u y b

19. Why We Need Multi Layer ?
Linear Separable:
Linear inseparable:
Solution?
yx yx
yx

20. k
O
jkw
Output nodes
Input nodes
Hidden nodes
Output Class
Input Record : x
i
w
ij- weights
Network is fully connected
j
O
A Multilayer Feed-Forward
Neural Network

21. Neural Network Learning
The inputs are fed simultaneously into the
input layer.
The weighted outputs of these units are fed
into hidden layer.
The weighted outputs of the last hidden layer
are inputs to units making up the output layer.

22. A Multilayer Feed Forward Network
The units in the hidden layers and output layer are
sometimes referred to as neurodes, due to their
symbolic biological basis, or as output units.
A network containing two hidden layers is called a
three-layer neural network, and so on.
The network is feed-forward in that none of the
weights cycles back to an input unit or to an output
unit of a previous layer.

23. A Multilayered Feed – Forward Network
INPUT: records without class attribute with
normalized attributes values.
INPUT VECTOR: X = { x1, x2, …. xn}
where n is the number of (non class) attributes.
INPUT LAYER – there are as many nodes as non-
class attributes i.e. as the length of the input vector.
HIDDEN LAYER – the number of nodes in the
hidden layer and the number of hidden layers
depends on implementation.

24. A Multilayered Feed–Forward
Network
OUTPUT LAYER – corresponds to the class attribute.
 There are as many nodes as classes (values of the
class attribute).
k
O
k= 1, 2,.. #classes
• Network is fully connected, i.e. each unit provides input
to each unit in the next forward layer.

25. Classification by Back propagation
Back Propagation learns by iteratively
processing a set of training data (samples).

For each sample, weights are modified to
minimize the error between network’s
classification and actual classification.

26. Steps in Back propagation
Algorithm
STEP ONE: initialize the weights and biases.
The weights in the network are initialized to
random numbers from the interval [-1,1].
 Each unit has a BIAS associated with it
The biases are similarly initialized to random
numbers from the interval [-1,1].
STEP TWO: feed the training sample.

27. Steps in Back propagation Algorithm
( cont..)
STEP THREE: Propagate the inputs forward;
we compute the net input and output of each
unit in the hidden and output layers.
STEP FOUR: back propagate the error.
STEP FIVE: update weights and biases to
reflect the propagated errors.
STEP SIX: terminating conditions.

28. Propagation through Hidden
Layer ( One Node )
The inputs to unit j are outputs from the previous layer. These are
multiplied by their corresponding weights in order to form a
weighted sum, which is added to the bias associated with unit j.

A nonlinear activation function f is applied to the net input.
-
f
weighted
sum
Input
vector x
output y
Activation
function
weight
vector
w

w
0j
w
1j
w
nj
x
0
x
1
x
n
Bias j

29. Propagate the inputs forward
For unit j in the input layer, its output is equal to its
input, that is,
jj
IO
for input unit j.
• The net input to each unit in the hidden and output
layers is computed as follows.
•Given a unit j in a hidden or output layer, the net input is
 
i
jiijj
OwI 
where wij is the weight of the connection from unit i in the previous layer to
unit j; Oi is the output of unit I from the previous layer;
j
is the bias of the unit

30. Propagate the inputs forward
Each unit in the hidden and output layers takes its net input
and then applies an activation function. The function
symbolizes the activation of the neuron represented by the
unit. It is also called a logistic, sigmoid, or squashing function.
Given a net input Ij to unit j, then
Oj = f(Ij),
the output of unit j, is computed as
jIj
e
O



1
1

31. Back propagate the error
When reaching the Output layer, the error is
computed and propagated backwards.
For a unit k in the output layer the error is
computed by a formula:
))(1(
kkkkk OTOOErr 
•
Where O k – actual output of unit k ( computed by activation
function.
Tk – True output based of known class label; classification of
training sample
Ok(1-Ok) – is a Derivative ( rate of change ) of activation function.
kIk
e
O



1
1

32. Back propagate the error
The error is propagated backwards by updating weights
and biases to reflect the error of the network classification .
For a unit j in the hidden layer the error is computed by a
formula:
•
jk
k
kjjj wErrOOErr  )1(
where wjk is the weight of the connection from unit j to unit
k in the next higher layer, and Errk is the error of unit k.

33. Update weights and biases
Weights are updated by the following equations, where l is a constant between 0.0
and 1.0 reflecting the learning rate, this learning rate is fixed for implementation.
ijij OErrlw )(
ijijij
www 
• Biases are updated by the following equations
jjj  
jj
Errl)(

34. Update weights and biases
We are updating weights and biases after the
presentation of each sample.
This is called case updating.
Epoch --- One iteration through the training set is called an
epoch.
Epoch updating ------------
Alternatively, the weight and bias increments could be
accumulated in variables and the weights and biases
updated after all of the samples of the training set have
been presented.
Case updating is more accurate

35. Terminating Conditions
Training stops
ij
w
• All in the previous epoch are below some
threshold, or
•The percentage of samples misclassified in the previous
epoch is below some threshold, or
• a pre specified number of epochs has expired.
• In practice, several hundreds of thousands of epochs may
be required before the weights will converge.

36. Output nodes
Input nodes
Hidden nodes
Output vector
Input vector: x
i
w
ij
 
i
jiijj
OwI 
))(1( kk
kkk
OTOOErr 
jk
k
kjjj
wErrOOErr  )1(
ijijij OErrlww )(
jjj
Errl)(
jIj
e
O



1
1
Backpropagation Formulas

37. Example of Back propagation
x1 x2 x3w14 w15 w24 w25 w34 w35 w46 w56
1 0 1 0.2 -0.3 0.4 0.1 -0.5 0.2-0.3-0.2
Initial Input and weight
Initialize weights :
Input = 3, Hidden
Neuron = 2 Output =1
Random Numbers
from -1.0 to 1.0

38. Example ( cont.. )
Bias added to Hidden
+ Output nodes
Initialize Bias
Random Values from
-1.0 to 1.0
Bias ( Random )
θ4 θ5 θ6
-0.4 0.2 0.1

39. Net Input and Output Calculation
Unitj Net Input Ij Output Oj
4
0.2 + 0 + 0.5 -0.4 = -0.7

5
-0.3 + 0 + 0.2 + 0.2 =0.1
6
(-0.3)0.332-(0.2)
(0.525)+0.1= -0.105
1.0
1
1



e
O
j
7.0
1
1
e
O
j


105.0
1
1
e
O
j


= 0.332
= 0.525
= 0.475

40. Calculation of Error at Each
Node
Unit j Error j
6

0.475(1-0.475)(1-0.475) =0.1311
We assume T 6 = 1
5
0.525 x (1- 0.525)x 0.1311x
(-0.2) = 0.0065
4
0.332 x (1-0.332) x 0.1311 x
(-0.3) = -0.0087

41. Calculation of weights and Bias
Updating
Learning Rate l =0.9
Weight New Values
w46
-0.3 + 0.9(0.1311)(0.332) = -
0.261
w56
-0.2 + (0.9)(0.1311)(0.525) = -
0.138
w14
0.2 + 0.9(-0.0087)(1) = 0.192
w15
-0.3 + (0.9)(-0.0065)(1) = -0.306
……..similarly ………similarly
θ6

0.1 +(0.9)(0.1311)=0.218
……..similarly ………similarly

42. Advanced Features of Neural
Network
Training with Subsets
Modular Neural Network
Evolution of Neural Network

43. Variants of Neural Networks
Learning
Supervised learning/Classification
•Control
•Function approximation
•Associative memory
Unsupervised learning or Clustering

44. Training with Subsets
Select subsets of data
Build new classifier on subset
Aggregate with previous classifiers
Compare error after adding classifier
Repeat as long as error decreases

45. Training with subsets
Subset 1
Subset 2
Subset 3
Subset n
NN 1
NN 2
NN 3
NN n
A Single
Neural Network
Model
The
Whole
Datase
t
Split the dataset
into subsets
that can fit
into memory
.
.
.

46. Modular Neural Network
Modular Neural Network
•Made up of a combination of several neural
networks.
The idea is to reduce the load for each neural
network as opposed to trying to solve the
problem on a single neural network.

47. Evolving Network Architectures
Small networks without a hidden layer can’t
solve problems such as XOR, that are not
linearly separable.
•Large networks can easily overfit a problem
to match the training data, limiting their
ability to generalize a problem set.

48. Constructive vs Destructive
Algorithm
Constructive algorithms take a minimal
network and build up new layers nodes and
connections during training.
Destructive algorithms take a maximal
network and prunes unnecessary layers
nodes and connections during training.

49. Training Process of the MLP
The training will be continued until the RMS is
minimized.
Global Minimum
Local Minimum
Local Minimum
ERROR
W (N dimensional)

50. Faster Convergence
Back prop requires many epochs to converge
Some ideas to overcome this
•Stochastic learning
•Update weights after each training example
•Momentum
•Add fraction of previous update to current update
•Faster convergence

51. Applications-I
Handwritten Digit Recognition
Face recognition
Time series prediction
Process identification
Process control
Optical character recognition

52. Application-II
Forecasting/Market Prediction: finance and banking
 Manufacturing: quality control, fault diagnosis
Medicine: analysis of electrocardiogram data, RNA & DNA
sequencing, drug development without animal testing

 Control: process, robotics

53. Summary
We presented mainly the followings-------
Basic building block of Artificial Neural Network.
Construction , working and limitation of single layer neural
network (Single Layer Neural Network).
Back propagation algorithm for multi layer feed forward NN.
Some Advanced Features like training with subsets, Quicker
convergence, Modular Neural Network, Evolution of NN.
Application of Neural Network.

54. Remember…..
ANNs perform well, generally better with larger number of
hidden units
More hidden units generally produce lower error
Determining network topology is difficult
Choosing single learning rate impossible
Difficult to reduce training time by altering the network topology
or learning parameters
NN(Subset) often produce better results

55. Question ???
Questions and Comments are welcome…

?
 THANKS
 Have a great Day !